|
Download research/agentic-audit/2026-10-10/acceptance/github-notebook/README.md from Oracle4CEG/OracleEconLab-Oracle-Incentives-v1: direct link, hf CLI and curl.
- Browser
- Download file 2.41 kB
-
https://huggingface.co/datasets/Oracle4CEG/OracleEconLab-Oracle-Incentives-v1/resolve/main/research/agentic-audit/2026-10-10/acceptance/github-notebook/README.md
- Command line
-
hf download hf://datasets/Oracle4CEG/OracleEconLab-Oracle-Incentives-v1/research/agentic-audit/2026-10-10/acceptance/github-notebook/README.md
-
curl -L -o README.md https://huggingface.co/datasets/Oracle4CEG/OracleEconLab-Oracle-Incentives-v1/resolve/main/research/agentic-audit/2026-10-10/acceptance/github-notebook/README.md
2.41 kB
| # Recorded public notebook execution | |
| These are actual outputs from [GitHub Actions run 38053135664](https://github.com/virusLuke3/Oracle-Agentic-Audit/actions/runs/38053135664) | |
| on 10 October 2026, not an unexecuted example notebook. The workflow ran all | |
| 6 code cells in a fresh environment, downloaded the public inputs | |
| without authentication, ran 32 tests, refitted CPU baselines and recomputed results. | |
| Notebook elapsed time was 762.12 seconds, including setup and downloads. | |
| - Notebook source: `9edb0de1e1a28f98b5bc38d4c59e55cf679fb2fe`. | |
| - Scientific implementation: `33ec7020e97fc6b8dd0005415d3184ce889d3e28`. | |
| - Input dataset: `ee197fc6fd6b8d980e7c7dda4feae966efe5d28c`. | |
| - 73 scientific CSV comparisons passed with absolute/relative tolerance 1e-10. | |
| - 18 numeric LaTeX tables and the exact prompt listing matched byte for byte. | |
| - Eight figures were regenerated; 0/8 PNGs matched the reference bytes. | |
| Numerical comparisons passed; pixel-identical rendering is not claimed. | |
| - No LLM inference requests were made. Saved response replay and new generation | |
| are different reproducibility claims. | |
| - A real hosted Google Colab run remains **NOT_RUN**. | |
| `acceptance.json` summarizes the execution. `research.executed.ipynb` preserves all | |
| actual cell outputs. `comparison_receipt.json` provides per-output checks. | |
| `download_receipt.json` records archive/file verification. `validation-output.txt` | |
| contains the real test and required-panel output. `environment.freeze.txt` records | |
| the scientific environment; notebook orchestration dependencies are pinned in | |
| `requirements-notebook.txt`. | |
| `artifact-verification.json` binds these downloaded records to GitHub artifact | |
| 11670288235 and its SHA-256 digest. `notebook-ci.json` preserves job/step status. | |
| `delivery-attempts.json` separately records earlier local setup/download interruptions, | |
| which did not reach scientific analysis. They are not relabeled successful runs. | |
| The generated data, tables and figures are in `generated-results.tar.gz` on Hugging | |
| Face under `research/agentic-audit/2026-10-10/acceptance/github-notebook/`. | |
| That large archive is deliberately excluded from Git; its checksum is in the | |
| artifact-verification receipt. Original model/data artifacts remain in the separate, | |
| immutable input release. This verification does not clear the complete Atlas, | |
| certify unknown public model revision/training metadata, or replace scientific review. | |